Claude Code v2.1.90 Ships /powerup: 18-Lesson In-Terminal AI Tutorials

Claude Code Gets Its First Built-In Tutorial System

On April 1, 2026, Anthropic shipped Claude Code v2.1.90, introducing the /powerup command—the first official, first-party interactive tutorial system built directly into the terminal-based AI coding agent. Over the following three days, versions v2.1.91 and v2.1.92 extended the update with enterprise security controls, expanded MCP result persistence, and a 60% performance improvement to the Write tool.

The release marks a shift in how Anthropic is approaching developer onboarding for Claude Code. Rather than relying on external documentation alone, the team has embedded structured, animated lessons directly into the terminal workflow where developers already spend their time.

/powerup: 18 Lessons Across Three Levels

What It Is

/powerup is a terminal-native curriculum built using the same React + Ink framework that powers the rest of the Claude Code interface. It delivers 18 structured lessons organized into three difficulty levels, each including a conceptual explanation, an animated demonstration running in-terminal, and a guided hands-on exercise.

Curriculum Overview

Beginner (6 lessons): CLAUDE.md configuration, the /clear command, /compact for context management, Plan Mode, and Model Selection. These lessons establish the foundational workflows that make Claude Code effective for everyday use.

Intermediate (6 lessons): Skills, Hooks, Sub-agents, MCP configuration, and the /rewind feature. This tier covers the orchestration and customization layer that transforms Claude Code from a chat tool into an agentic workflow engine.

Advanced (6 lessons): Worktrees for parallel workstreams, Auto Mode, /schedule for background tasks, and SDK Headless Mode for programmatic control. The advanced tier targets developers building automation pipelines and CI/CD integrations on top of Claude Code.

Design Philosophy

Each lesson is self-contained and can be completed in under five minutes. The animated demos run in the same terminal window, meaning developers can immediately replicate what they see without switching context to a browser. Anthropic's decision to build this system in-terminal reflects a broader principle in Claude Code's UX: keep the developer in flow.

v2.1.91: MCP Result Persistence

Released on April 3, Claude Code v2.1.91 raised the MCP tool result size limit to 500,000 characters via the anthropic/maxResultSizeChars metadata field. Previously, large tool results—such as complete database schemas, codebase indexes, or long API responses—would be truncated, forcing developers to work around the limit manually.

The update also allows plugins to ship executable files in a bin/ directory, enabling custom CLI tools to be invoked directly by Claude during agentic sessions without requiring external PATH configuration.

v2.1.92: Enterprise Controls and Performance

On April 4, v2.1.92 added the forceRemoteSettingsRefresh policy, which forces Claude Code to pull fresh configuration from the organization's remote settings server on every startup. This addresses a pain point for enterprise teams managing Claude Code across large developer populations, where configuration drift between local and organizational policies had created compliance inconsistencies.

On the performance side, the Write tool's diff computation speed for large files was improved by approximately 60%. The fix specifically targets files containing tabs, ampersands, and dollar signs—characters that previously caused quadratic-time processing on large buffers.

Usability Analysis

For developers new to Claude Code, /powerup lowers the learning curve significantly. Many of Claude Code's most powerful features—Hooks, Sub-agents, MCP—have previously required reading through external documentation before they could be used effectively. The in-terminal tutorial format removes that friction.

For experienced users, the advanced lessons on Worktrees and SDK Headless Mode offer a structured path into capabilities that are often underutilized. The curriculum is dense enough to provide value even to developers who have been using Claude Code for months.

The enterprise controls in v2.1.92 will be particularly relevant for organizations deploying Claude Code at scale. Consistent configuration enforcement has been a recurring concern as teams grew, and forceRemoteSettingsRefresh directly addresses it.

Pros and Cons

Pros:

• In-terminal tutorial format reduces context-switching during onboarding
• Three-tier curriculum accommodates beginner through advanced users
• MCP result persistence at 500k characters eliminates truncation for large data payloads
• 60% Write tool speed improvement is meaningful for large-codebase workflows
• Enterprise configuration policy provides better compliance control

Cons:

• /tag and /vim commands were removed in v2.1.92, which may affect users who relied on them
• The tutorial system requires active participation—passive reading is not supported
• Advanced lessons assume familiarity with agentic concepts that may overwhelm newcomers who rush through levels

Outlook

The /powerup system positions Claude Code to grow its developer base without proportional increases in support overhead. As Claude Code expands into enterprise and team environments—where new users are onboarded constantly—an embedded curriculum that runs in the same tool is a scalable approach to education.

The MCP improvements in v2.1.91 are equally significant for the broader agentic ecosystem. Larger tool result payloads mean Claude Code can now ingest and act on more complex real-world data sources without truncation workarounds, expanding the range of automation tasks it can handle reliably.

Conclusion

Claude Code v2.1.90 through v2.1.92 represent a coherent product push: better onboarding via /powerup, stronger agentic data handling via MCP persistence, and improved enterprise governance via configuration enforcement. Developers already using Claude Code should update immediately, particularly for the Write tool performance gains. New users should start with /powerup before diving into the documentation—it is the fastest path to effective use.